Traitlet-based interaction with the jupyter widget is clunky and incomplete

[ceriottm]

There's a lot of potential in the interaction between the chemiscope widget and a notebook interface.
Currently this is mostly based on exposing a traitlet interface for settings. There are however many issues and features that could be added (relatively) easily:

• The only way to change a setting is to set the whole settings traitlet, as changing one of the dict items does not trigger the traitlet update. This is annoying but probably unavoidable. Worth thinking if a better API is possible though
• Errors when setting an invalid dict are clunky: they are shown as chemiscope errors rather than propagated to the point where the dict is set, and only occur when using an invalid dict key. Invalid settings values are ignored, which should at the very least be fixed
• The traitlet interface is largely undocumented. Explaining how to use it and having an example would be great for adoption.
• There are many other modes of interaction that could be enabled. A traitlet interface to control the selected point is an obvious one (addressed partly by this defunct pr Python-side selection #316) but also controlling the orientation and zoom of the structure or the map region might have useful use cases.